Posted on 10/20/2008 2:12:41 PM PDT by Chet 99
Hmm....a nine point lead with a seven percent Dem oversample? Without doing the math, I’m guessing that this translates into a small national lead for McCain. To put this in context, it took a 4 point Dem turnout advantage to give Gore a 0.5% margin in the 2000 popular vote. This happens because the past election history shows that slightly more Democrats vote Republican than vice-versa.
If you were designing a sample frame, how would weight your sample between Dem and Rep?
When comparing your survey data to Census Tract data how much of a disparity between the two would you believe to be acceptable?
And when conducting your poll what response rate would you aim for?
Historically, the two parties are within 1% or 2% after the votes are counted, never more than 4%. There is no basis in actual voting statistics for a 7% or 12% difference.
You caught me. I have no idea how pollsters go about developing the list of names to call. But it doesn’t matter nearly as much as the deal you’re making of it.
It matters which calls you ‘throw out’ to get down to your weighted numbers. And that’s where I’m saying they’re under-sampling Republicans. I’m saying the pollsters are not giving sufficient weight to the historical relative percentages.
Of course, you’ll say they’re the pros and I’m just an anonymous poster on FR. True.
Even so, we won’t know the real truth until sometime after Nov 4.
The current weighting of R & D is based on how people have Id’ed themselves over the past six months. Now, why is it that so many more people identified themselves as Democrats in the last six months?
However, the sample frame is based on census data and when a bogus survey is released it is easy to debunk by comparing the frame to census data.
We’ll see.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.